Twilio expands CustomerAI capabilities with generative and predictive AI

At its annual Signal conference today, Twilio is announcing a significant expansion of its customer artificial intelligence (AI) tools, dubbed CustomerAI.

Twilio has been steadily building out its partnerships and technologies for AI over the course of 2023 in the lead-up to Signal. Earlier this month the company announced a partnership with OpenAI as a precursor to the larger set of CustomerAI announcements being made today. 

At Signal, Twilio is expanding its CustomerAI capabilities across its product portfolio which includes customer data platform (CDP), contact center and marketing capabilities.  Among the new features is voice intelligence, which pulls insights from conversations, predictive analytics and generative journeys for building marketing campaigns.

Twilio is also aiming to enable responsible and explainable AI usage through the use of its AI Nutrition Facts Labels, which will provide a bill of materials for AI models used for a specific service.

Looking beyond just AI, Twilio is also improving the way it organizes and shares profiles, thanks in part to a partnership with Databricks that has Twilio using the Databricks Delta Lake data lakehouse and Delta Sharing technologies.

“Customer AI is both predictive and generative,” Kathryn Murphy, SVP of Product at Twilio told VentureBeat. “CustomerAI is really getting us the technology and tools to create this flywheel of understanding using the data that we gain and understanding for better engagement.”

CustomerAI bringing new intelligence to voice conversations

One of the new features coming from Twilio is a capability the company is calling voice intelligence.

Murphy explained that voice intelligence uses the power of large language models (LLMs) to understand conversations and extracts traits from those conversations. Traits could be any number of different attributes that an organization might want to keep track of, such as a user’s preferences as expressed in a conversation.

Those traits can be inferred by the voice intelligence technology and then injected into the Twilio Segment customer data platform (CDP) to help enable a better overall customer experience.

The voice intelligence technology is built using multiple LLM models including some that have been built by Twillio, as well as leveraging technology from OpenAI.

Predictive and generative AI enable Twilio’s CustomerAI

CustomerAI Predictions is one of the predictive capabilities that Twilio is announcing as generally available at Signal.

Murphy said that CustomerAI Predictions uses customer data and events to build individual models for each customer that can then be used to generate predictions about things like customer lifetime value, likelihood to purchase and customer churn. These predictions can be surfaced in applications to improve conversion.

Those predictions can also be used to fuel the new Twilio Generative Journeys capability announced today.

“Generative Journeys is interesting because it’s leveraging the predictions to then build a campaign,” said Murphy.

She explained that with Generative Journeys, a marketer can express with natural language what they want to achieve, for example saying they want to win back customers that haven’t visited the company’s website in the last six months. Previously, a marketer had to go through a marketing journey builder approach, selecting the audience and figuring out next steps.

“Generative Journeys actually does all of that for the marketer and takes into account all the different predictions and traits from the customers in that audience to pick the right number of steps,” she said.

Linked Profiles flow from the data lakehouse

Beyond just AI updates, Twilio is also out with a data-related update for a feature called Linked Profiles that benefits from Twilio’s partnership with data lakehouse provider Databricks.

Linked Profiles allows organizations to model relationships between customer profiles at a household or account level. Murphy explained that the new feature is important for understanding buying groups in complex B2B or B2C scenarios. She noted that Linked Profiles could track events like different family members watching different movies and then market to the whole household. For B2B, it helps understand the account and buying team members.

With the Databricks data lakehouse, Murphy said that Twilio is able to use a ‘zero copy architecture’ where data can be shared to create a linked profile, without the need to copy the same data into multiple locations.

Looking forward, Murphy said that there is more potential opportunity for Twilio to work with Databricks, particularly on AI efforts. Databricks acquired MosaicML for $1.3 billion in June to expand its AI capabilities.

“We’re pretty excited about their Mosaic acquisition as that just adds another level,” she said.